asr_mind_mode

This model is a fine-tuned version of w11wo/wav2vec2-xls-r-300m-korean on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.4669
• Wer: 1.1260

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5
• training_steps: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
7.4112	0.1667	5	2.5890	1.0324
6.8732	0.3333	10	2.2256	1.0513
7.6868	0.5	15	2.0113	1.0675
7.5304	0.6667	20	1.8841	1.0846
6.0630	0.8333	25	1.6901	1.0999
6.2924	1.0	30	1.5908	1.1152
6.0157	1.1667	35	1.5801	1.1215
5.2062	1.3333	40	1.5709	1.1251
5.4203	1.5	45	1.5565	1.1260
4.7041	1.6667	50	1.5526	1.1242
5.3392	1.8333	55	1.5531	1.1251
6.4507	2.0	60	1.5481	1.1251
5.8050	2.1667	65	1.5197	1.1269
4.5138	2.3333	70	1.5034	1.1287
5.0324	2.5	75	1.4949	1.1278
5.8347	2.6667	80	1.4905	1.1260
5.9266	2.8333	85	1.4765	1.1242
5.3955	3.0	90	1.4727	1.1251
5.0815	3.1667	95	1.4682	1.1242
5.5879	3.3333	100	1.4669	1.1260

Framework versions

• Transformers 5.15.1
• Pytorch 2.11.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2